Am 06.02.2010 03:45, schrieb Sujit V:
> I have integrated the kdump in our linux 2.6.23 based kernel. If I do
> echo c > /proc/sysrq-trigger then it boots the kdump kernel & I use
> the cp /proc/vmcore /local/crash/vmcore-incomplete
>
> On a new x86 based hardware (32 bit PAE kernel) the "cp /proc/vmcore
> " command exits prematurely.
>
> cp -v /proc/vmcore /local/crash/vmcore-incomplete
> cp: reading `/proc/vmcore': Invalid argument >>> Error >>
>
> ls -l /local/crash/
> total 115588
> -r-------- 1 0 0 2526724096 Feb 5 03:09 vmcore-incomplete
>
> ls -l /proc/vmcore
> -r-------- 1 0 0 4153763584 Feb 5 03:16 /proc/vmcore
>
>
> Every time I trigger kdump the cp command exits after copying
> 2526724096 bytes. So I thought it might be a file size issue.
> Googling pointed out a patch in include/linux/proc_fs.h
I would compare 'readelf -l /proc/vmcore' (ELF program headers) and map
the file offset 2526724096 to the physical memory that belongs to that
file offset. Maybe accessing that physical memory fails for some reason.
> struct vmcore {
> struct list_head list;
> unsigned long long paddr;
> - unsigned long size;
> + unsigned long long size;
> loff_t offset;
>
> I checked my kernel & it already has this patch.
I don't think that this is the problem because 2526724096 is between 2G
and 4G, so it doesn't point to some "magic border".
Regards,
Bernhard
